agentic-ai-engineering/site/.vitepress/dist/assets/labs_l7-autoresearch.md.BYlzPLYo.js

import{c as i,Q as a,j as t,m as e}from"./chunks/framework.BPKcPtvA.js";const E=JSON.parse('{"title":"L7a: Autoresearch Loop","description":"","frontmatter":{},"headers":[],"relativePath":"labs/l7-autoresearch.md","filePath":"labs/l7-autoresearch.md","lastUpdated":1780488246000}'),h={name:"labs/l7-autoresearch.md"};function n(l,s,k,r,p,o){return a(),t("div",null,[...s[0]||(s[0]=[e(`<h1 id="l7a-autoresearch-loop" tabindex="-1">L7a: Autoresearch Loop <a class="header-anchor" href="#l7a-autoresearch-loop" aria-label="Permalink to &quot;L7a: Autoresearch Loop&quot;">​</a></h1><p>Build a self-improving agent with integrity guards.</p><p><strong>Module</strong>: M7 Advanced Topics<br><strong>Est. Time</strong>: 75 min<br><strong>Files</strong>: <code>starter.py</code>, <code>solution.py</code></p><h2 id="objective" tabindex="-1">Objective <a class="header-anchor" href="#objective" aria-label="Permalink to &quot;Objective&quot;">​</a></h2><p>Create an agent that runs experiments, measures its own performance, logs results, and decides whether to keep or discard each change.</p><h2 id="concepts" tabindex="-1">Concepts <a class="header-anchor" href="#concepts" aria-label="Permalink to &quot;Concepts&quot;">​</a></h2><ul><li>Autoresearch: run → measure → log → decide</li><li>Code hashing to detect grinding</li><li>Median vs best comparison</li><li>Reward hack defense</li></ul><h2 id="starter" tabindex="-1">Starter <a class="header-anchor" href="#starter" aria-label="Permalink to &quot;Starter&quot;">​</a></h2><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">cd</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> course/labs/L7-autoresearch/</span></span>
<span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">python</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> starter.py</span></span></code></pre></div><h2 id="solution" tabindex="-1">Solution <a class="header-anchor" href="#solution" aria-label="Permalink to &quot;Solution&quot;">​</a></h2><div class="language-bash vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">bash</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#6F42C1;--shiki-dark:#B392F0;">python</span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;"> solution.py</span></span></code></pre></div><h2 id="experiment-log-format" tabindex="-1">Experiment Log Format <a class="header-anchor" href="#experiment-log-format" aria-label="Permalink to &quot;Experiment Log Format&quot;">​</a></h2><div class="language-jsonl vp-adaptive-theme"><button title="Copy Code" class="copy"></button><span class="lang">jsonl</span><pre class="shiki shiki-themes github-light github-dark vp-code" tabindex="0"><code><span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">{</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;run&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">1</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;status&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;baseline&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;metric&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: {</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;name&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;latency&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;value&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">52</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;unit&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;ms&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}}</span></span>
<span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">{</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;run&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">2</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;status&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;keep&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;metric&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: {</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;name&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;latency&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;value&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">46</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;deltaPct&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">-11.5</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}</span></span>
<span class="line"><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">{</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;run&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">3</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;status&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;discard&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;metric&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: {</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;name&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#032F62;--shiki-dark:#9ECBFF;">&quot;latency&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;value&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">53</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}, </span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">&quot;deltaPct&quot;</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">: </span><span style="--shiki-light:#B31D28;--shiki-light-font-style:italic;--shiki-dark:#FDAEB7;--shiki-dark-font-style:italic;">+</span><span style="--shiki-light:#005CC5;--shiki-dark:#79B8FF;">1.9</span><span style="--shiki-light:#24292E;--shiki-dark:#E1E4E8;">}</span></span></code></pre></div><h2 id="integrity-guards" tabindex="-1">Integrity Guards <a class="header-anchor" href="#integrity-guards" aria-label="Permalink to &quot;Integrity Guards&quot;">​</a></h2><table tabindex="0"><thead><tr><th>Threat</th><th>Detection</th><th>Prevention</th></tr></thead><tbody><tr><td>Grinding (same code re-run)</td><td>Code hash comparison</td><td>Skip run</td></tr><tr><td>Noise-chasing</td><td>Median vs best comparison</td><td>Use median, not best</td></tr><tr><td>Reward hacking</td><td>Timing function isolation</td><td>Verify computation not shortcut</td></tr><tr><td>Test set leakage</td><td>Test data hash verification</td><td>Assert data unchanged</td></tr></tbody></table>`,15)])])}const u=i(h,[["render",n]]);export{E as __pageData,u as default};